Corpus Design Research Articles

Crossroads Corpus creation: Design and case study

Abstract This paper illustrates a methodological approach to the design of an annotated corpus using a case study of phonetic convergences and divergences by multilingual speakers in southwestern Senegal’s Casamance region. The newly compiled corpus contains approximately 183,000 annotations of multilingual, spoken data, gathered by eight researchers over a ten year span using methods ranging from structured lexical elicitation in controlled contexts to naturally occurring, multilingual conversations. The area from which the data were collected consists of three villages and their primary languages, and yet many more contribute to the linguistic landscape. Detailed metadata inform analyses of variation, the context in which a speech act took place and between whom, the speakers’ linguistic repertoires, trajectories, and social networks, as well as the larger language context. A potential path for convergence or divergence that emerged during data collection and in building and searching the corpus is the crossroads in the phonetic production of word-initial velar plosives. Word-initial [k] emerges in one language where only [ɡ] is present in the other; the third utilizes both. The corpus design makes it feasible, not only to identify areas of accommodation, but to grasp the context, enabling a sociolinguistically informed analysis of the speakers’ linguistic behavior.

Corpus-Based Training to Build Translation Competences and Translators’ Self-Reliance

Abstract Among well-established standard operating strategies and methodologies aimed at improving students’ multi-layered translation competence, corpus linguistics has given impulse to new interdisciplinary standpoints. Plugging in cutting edge toolkits, corpus use envisages real-life translation training in compliance with current market demands. The paper underpins specialised corpus design, using MAXQDA as an interactive tool meant to build translators’ functional autonomy.

‘Keep out of reach of children!’ Introducing the Corpus of Product Information (CoPI) and its potential for corpus-based genre teaching

In this paper, we introduce the language-pedagogic potential of the Corpus of Product Information (CoPI). The corpus is XML-annotated and contains about 100,000 words of product descriptions of health products, cleaning supplies and products for beauty and personal care, divided into three textual moves: (1) overview, (2) directions and (3) warnings. First, we describe the data collection, corpus design and annotation scheme of the corpus, and then we present the findings of an analysis of CoPI's most frequent words, clusters and its type–token ratio. Finally, we show its potential for language-pedagogic purposes and suggest how the CoPI analyses can be used for paper- and computer-based DDL activities that foster corpus-based genre teaching in the advanced EFL classroom. We conclude this paper by summarising the outcomes of a first case study we conducted to test these activities with advanced learners of English.

장르 기반 영어 학습자 코퍼스의 이메일 오류 분석: 학부 교양영어 교육을 중심으로

As a preliminary study for the effective development of a genre-focused English learner corpus, this article aims to investigate most frequent error types and their frequencies in English emails written by undergraduate freshmen in South Korea. Data for this study include English emails of 86 Korean students majoring in humanities or social science in a university located in Daegu. With the rise of Internet, ESL/EFL education has witnessed a growing interest in teaching email usage in composition courses, as it provides a variety of opportunities to evaluate language abilities including interpersonal and pragmatic abilities. The present article revealed that the most frequent error type was concerned with style, such as capitalization (28.7%) and punctuation (7.8%), which was followed by determiner deletion (6.3%), genre convention such as closing (3%), countability of nouns (3%), and verb choice (2.7%). Different error types and frequencies were identified according to different English proficiency levels (Korean SAT and TOEIC), which evidenced the need to include the English proficiency level annotation in the corpus design and to focus on different types of errors in class in accordance with learners’ proficiency levels.

Metodología de elaboración de un glosario bilingüe y bidireccional (inglés-español/español-inglés) basado en corpus para la traducción de manuales de instrucciones de televisores

Abstract The sources of information that translators may use are extremely varied, ranging from oral consultation with an expert to a search using specialised dictionaries and glossaries. Nowadays, however, one the most relevant documentation activities in the field of Translation involves the use of Internet resources and, closely related to this, the compilation and management of virtual corpora. For this reason, in the present paper we present a systematic methodology for extracting bilingual and bidirectional glossaries (English-Spanish/Spanish-English) based on parallel corpora to translate TV User Manuals. In fact, according to art. 5 of the Council Resolution of 17 December 1998 on operating instructions for technical consumer goods (98/C 411/01) it is essential to control the quality when writing and translating these manuals. In order to illustrate this methodology we focus on corpus design (according to the skopo) and on the compilation protocol (in four steps: searching, downloading, text formatting and saving data) in order to ensure quality. As for the quantity, we check the quantitative representativeness with the ReCor software (cfr. Seghiri 2006: 387). Once the corpus is representative from the qualitative and the quantitative points of view, it can be managed with a concordance program. So, we illustrate how to extract the terms semiautomatically in order to build a bilingual and bidirectional glossary with a parallel concordance named ParaConc. Thus, in the present paper we combine the main resource for researchers (cfr. Bowker 1998; Varantola 2000; Seghiri 2011) within the Translation field: corpora, in order to ensure quality; and the main documentation resource for prospective translators (cfr. Corpas et al. 2001): bilingual glossaries.

Sentence Selection Based on Extended Entropy Using Phonetic and Prosodic Contexts for Statistical Parametric Speech Synthesis

This paper proposes a sentence selection technique for constructing phonetically and prosodically balanced compact recording scripts for speech synthesis. In the conventional corpus design of speech synthesis, a greedy algorithm that maximizes phonetic coverage is often used. However, for statistical parametric speech synthesis, balances of multiple phonetic and prosodic contextual factors are important as well as the coverage. To take account of both of the phonetic and prosodic contextual balances in sentence selection, we introduce an extended entropy of phonetic and prosodic contexts, such as biphone/triphone, accent/stress/tone, and sentence length. For detailed investigation, conventional and proposed techniques are evaluated using Japanese, English, and Chinese corpora. The objective experimental results show that the proposed technique achieves better coverage and balance of contexts. In addition, speech synthesis experiments based on hidden Markov models reveal that the generated speech parameters become closer to those of the natural speech compared with other conventional sentence selection techniques. Subjective evaluations show that the proposed sentence selection based on the extended entropy improves the naturalness of the synthetic speech while maintaining the similarity to the original sample.

The Influence of Annotation, Corpus Design, and Evaluation on the Outcome of Automatic Classification of Human Emotions

The integration of emotions into human computer interaction applications promises a more natural dialog between the user and the technical system he operates. In order to construct such machinery, continuous measuring of the affective state of the user becomes essential. While basic research that is aimed to capture and classify affective signals has progressed, many issues are still prevailing that hinder easy integration of affective signals into human-computer interaction. In this paper, we identify and investigate pitfalls in three steps of the work-flow of affective classification studies. It starts with the process of collecting affective data for the purpose of training suitable classifiers. Emotional data has to be created in which the target emotions are present. Therefore, human participants have to be stimulated suitably. We discuss the nature of these stimuli, their relevance to human-computer interaction and the repeatability of the data recording setting. Second, aspects of annotation procedures are investigated, which include the variances of individual raters, annotation delay, the impact of the used annotation tool and how individual ratings are combined to a unified label. Finally, the evaluation protocol is examined which includes, amongst others, the impact of the performance measure on the accuracy of a classification model. We hereby focus especially on the evaluation of classifier outputs against continuously annotated dimensions. Alongside the discussed problems and pitfalls and the ways how they affect the outcome, we provide solutions and alternatives to overcome these issues. As a final part of the paper we sketch a recording scenario and a set of supporting technologies that can contribute to solve many of the issues mentioned above.

Compiling computer-mediated spoken language corpora

This paper discusses key issues in the compilation of spoken language corpora in a computer-mediated communication (CMC) environment, using data from the Corpus of Academic Spoken English (CASE), a corpus ofSkypeconversations currently being compiled at Saarland University, Germany, in cooperation with European and US partners. Based on first findings,Skypeis presented as a suitable tool for collecting informal spoken data. In addition, new recommendations concerning data compilation and transcription are put forward to supplement existing best practice as presented in Wynne (2005). We recommend the preservation of multimodal features during anonymisation, and the addition of annotation elements already at the transcription stage, particularly CMC-related discourse features, English as a Lingua Franca (ELF) features (e.g. non-standard language and code-switching), as well as the inclusion of prosodic, paralinguistic, and non-verbal annotation. Additionally, we propose a layered corpus design in order to allow researchers to focus on specific annotation features.

Corpus-Based Studies of Translational Chinese in English–Chinese Translation (2015). Richard Xiao and Xianyao Hu.

Corpus-Based Studies of Translational Chinese in English–Chinese Translation (2015). Richard Xiao and Xianyao Hu. Berlin, Heidelberg: Springer-Verlag. ISBN: 978-3-642-41362-9 Translation universals (TUs) have been one of the research foci in corpus translation studies ever since Baker initiated this line of research in 1993.One key research effort has been to test out the TU hypotheses by bringing in different languages and cultures into the picture (Laviosa, 2011). However, the Anglo-centric bias in the pursuit of TUs is obvious, and evidence is, by and large, limited within the European languages, English in particular. This book by Xiao and Hu (2015) provides a much needed perspective and directs their efforts towards investigating the linguistic features of translational Chinese by probing into two genetically different languages, Chinese and English. The readers should be delighted to find that on the one hand, ‘translated texts in Chinese share a number of common properties’ (ibid. p. 172) of the TUs as have been in previous studies (e.g. Baker, 1996), and on the other hand, the previously defined and discussed TUs may be too over-generalized or simplified to reveal the true features of translational language, as a result of which we may need more refined TUs. In their book, Xiao and Hu mainly seek to answer the question, that is, to what extent the features of translational Chinese are in line with previous studies of TU hypotheses, such as explicitation, simplification, normalization, under-representation, convergence, and source language (SL) shining through. For that purpose, a detailed analysis was made of the typical linguistic features of translational Chinese from macro-statistic to specific lexical and grammatical level using a comparable corpus design involving the Lancaster Corpus Mandarin Chinese (LCMC) and … suwenchao0617{at}126.com

NUWT: JAWI-SPECIFIC BUCKWALTER CORPUS FOR MALAY WORD TOKENIZATION

This paper describes the design and creation of a monolingual parallel corpus for the Malay language written in Jawi.This paper proposes a new corpus called the National University of Malaysia Word Tokenization (NUWT) corpora To the best of our knowledge, currently, there is no sufficiently comprehensive, well-designed standard corpus that is annotated and made available for the public for the Jawi script corpora.This corpus contains the Jawi-specific Buckwalter character code and can be used to evaluate the performance of word tokenization tasks, as well as further language processing.The objective of this work is to conform and standardize the corpora between similar characters in Jawi.It consists of three subcorporas with documents from different genres. The gathering and processing steps, as well as the definition of several evaluation tasks regarding the use of these corpora, are included in this paper.One of the important roles and fundamental tasks of the corpus, which is the tokenization, is also presented in this paper.The development of the Malay language tokenizer is based on the syntactic data compatibility of Malay words written in Jawi.A series of experiments were performed to validate the corpus and to fulfill the requirement of the Jawi script tokenizer with an average error rate of 0.020255.Based on this promising result, the token will be used for the disambiguation and unknown word resolution, such as out-of vocabulary (OOV) problem in the tagging process.

An electromagnetic articulometer study of tongue and lip troughs

Troughs, which are a discontinuity in anticipatory coarticulation (Perkell 1968), have been shown to occur in the tongue body in /ipi/ and lips in /usu/. An electromagnetic articulometer (EMA) study of a single subject reported previously showed that intraoral pressure during the consonant could not be solely responsible for the downwards tongue body movement observed during the labial consonants (/p, b, m, f, v/); results supported the hypothesis of a secondary tongue gesture during the consonant. Lip troughs during /s/ in rounded contexts were more ambiguous, possibly because of inconsistencies in corpus design. Six new subjects have been recorded using a WAVE EMA system and a revised corpus. Tongue trough magnitudes decreased in the order /f, v/ > /p, b/ > /m/, serving to disprove the aerodynamic hypothesis; they were deeper for long than short consonants (e.g., [iffi] > [ifi]), and converged to the same point in asymmetric vowel contexts. The same type of convergence during the consonant was observed ...

The GUM corpus: creating multilayer resources in the classroom

This paper presents the methodology, design principles and detailed evaluation of a new freely available multilayer corpus, collected and edited via classroom annotation using collaborative software. After briefly discussing corpus design for open, extensible corpora, five classroom annotation projects are presented, covering structural markup in TEI XML, multiple part of speech tagging, constituent and dependency parsing, information structural and coreference annotation, and Rhetorical Structure Theory analysis. Layers are inspected for annotation quality and together they coalesce to form a richly annotated corpus that can be used to study the interactions between different levels of linguistic description. The evaluation gives an indication of the expected quality of a corpus created by students with relatively little training. A multifactorial example study on lexical NP coreference likelihood is also presented, which illustrates some applications of the corpus. The results of this project show that high quality, richly annotated resources can be created effectively as part of a linguistics curriculum, opening new possibilities not just for research, but also for corpora in linguistics pedagogy.

¿Qué traducen los traductores económicos del alemán-español y español-alemán? Estudio basado en encuestas

The aim of this paper is to present the results of a survey relating to economic and business translation from German into Spanish and vice versa, the objective of which is to build a specialised corpus in German. This research was conducted within the framework of the COMENEGO Project (Spanish acronym for Multilingual Corpus of Business and Economics). The purpose of the project is to create a stable electronic corpus for translation practitioners from Spanish into various languages and vice versa. The pilot corpus includes Spanish, French and English components. As with the English corpus, the results of a survey on the practice of translation in the field of business and economics are a valuable source of information for compiling the German corpus. Moreover, not only may the results enable us to circumscribe the corpus design, they could also aid translator trainers in the preparation of specialised courses.

영상번역교육을 위한 코퍼스 설계

This paper explores requirements for multimodal corpus design under the premise that special corpus for English-Korean AVT translation training should be built. For this, firstly, it provides a critical overview of the usefulness of corpus-based translation learning and the uniqueness of AVT translation training. Next, it discusses requirements for screen translation corpus design which is able to take into account multiple dimensions of English-Korean AVT translation, introducing the procedure and outcome of sample corpus preparatory to the screen translation corpus. This paper argues criteria for determining the size, scope, and range of the corpus. It also proposes how data should be annotated to help AVT trainees develop traditional linguistic, communicative, cultural skills and improve paralinguistic and technical competence. The argument and proposal is given on the basis of the outcome of the sample corpus. The research result indicates that the mutimodal corpus can offer valid data in a systematic way for AVT translation training.

A Structure for Annotation and Ground-truthing of Urdu Handwritten Text Image Corpus

Over the last few decades, a large evolution has been made in the field of handwritten recognition. Material of handwritten documents is become less with current trends of digital electronics. However, for the investigation and research on a particular language a large volume of handwritten documents database is required. In this paper we describe our approach for development a large volume of Urdu handwritten text images Corpus on Urdu language. To make the database available in large field of Natural Language Processing we annotate database for each image and associate a XML based ground-truth Meta information to make it computer compatible as a linguistic resource. This paper focus on the some issue related with Corpus design and annotation such as data collection, writers selection, methodology of annotation etc.

Design and Implementation of an Online Corpus of Presentation Transcripts of TED Talks

This paper describes the TED Corpus Search Engine (TCSE), an online corpus system that searches transcripts and translations of over 1,800 TED Talks, allowing users to query surface text forms, lemmas, parts-of-speech, or their combinations. Videos are available for the matched text, making the system useful for both linguistic research and language education. Also discussed is the system's theoretical background. The usage-based model of language is widely reflected in TCSE's structures and features. With functionalities that retrieve linguistic exemplars efficiently and provide multimodal contextual data, TCSE is a corpus system designed in accordance with an existing theoretical framework.

Turning the Corpus into a Functional Component of the Dictionary: The Case of the Oenolex Wine Dictionary

Lexicographical corpora for dictionary making are normally used as an empirical base of the dictionary. They provide highly valuable lexical material, patterns, attestations of use, etc. However, once exploited by the lexicographer, they remain hidden to the user. The question thus is: Can a corpus become even more useful not only to the lexicographer, but also to the user? How can a corpus become a functional component of the dictionary? I will address these questions by outlining the methodological issues related to the design and exploitation of an oral corpus for the construction of Oenolex, a dictionary of wine tasting. In Oenolex, the corpus is not simply the empirical base of the dictionary: it is lexicographically structured so that it can be accessed by the user. Oenolex also shows how oral corpora can be an asset for specialised dictionary projects, particularly in domains in which terminology is subject to a high degree of variation. This in turn calls for an even closer cooperation with experts; I will therefore argue, on the basis of the Klosa model of lexicographical phases (2013), that experts from the field of specialised communication and knowledge covered by the dictionary are truly needed in almost all phases of the dictionary process.

Statistical Parametric Evaluation on New Corpus Design for Malay Speech Articulation Disorder Early Diagnosis

Speech-to-Text or always been known as speech recognition plays an important role nowadays especially in medical area specifically in speech impairment. In this study, a Malay language speech-to-Text system was been designed by using Hidden Markov Model (HMM) as a statistical engine with emphasizing the way of Malay speech corpus design specifically for Malay articulation speech disorder. This study also describes and tests the correct number of state to analyze the changes in the performance of current Malay speech recognition in term of recognition accuracy. Statistical parametric representation method was utilized in this study and the Malay corpus database was constructed to be balanced with all the phonetic placed and manner of articulation sample appeared in Malay speech articulation therapy. The results were achieved by conducting few experiments by collecting sample from 80 patient speakers (child and adult) and contain for almost 30,720 sample training data.

Tarpkalbiniai ir tarpdalykiniai mokslo kalbos tyrimai: medžiagos ir metodų pasirinkimo iššūkiai tyrėjams

Straipsnyje trumpai apžvelgiamos dominuojančios šiuolaikinių mokslo kalbos tyrimų kryptys, tačiau pagrindinis dėmesys skiriamas mokslinio straipsnio, kaip akademinio žanro, tyrimams ir jų metodikai. Mokslinio straipsnio struktūra ir jos ypatumai, autoriaus argumentavimo strategijos yra akademinio diskurso bruožai, sulaukę itin didelio kalbininkų susidomėjimo pasaulinėje lingvistikoje. Šio apžvalginio straipsnio tikslas – aptarti tarpdalykinių ir tarpkalbinių tyrimų specifiką ir išryškinti kai kuriuos metodologinius aspektus, svarbius sudarant specialiuosius palyginamuosius mokslo kalbos tekstynus. Remiantis naujausiais šios srities tyrimais, apžvelgiami pasaulio tyrėjų sprendimai tekstyno sandaros ir medžiagos analizės klausimais. Straipsnis galėtų būti naudingas studentams, jauniesiems tyrėjams, besidomintiems specialiųjų mokslo kalbos tekstynų sudarymu ir analize.

Evaluating reliability in quantitative vocabulary studies

Recent methodological advances have been used to create word lists based on large corpora. The present paper explores whether these corpora — and the associated lists — are unequivocally more representative. Corpus design considerations have usually focused on issues of external representativeness (representing the target discourse domain), while disregarding issues of internal representativeness (whether the corpus permits reliable descriptions of linguistic variation). This disregard may be especially problematic for studies of lexical variation, where it is difficult to achieve stable, reliable results from corpus analysis. The present paper illustrates these challenges through experiments based on analysis of a corpus representing a highly restricted discourse domain: university-level introductory psychology textbooks. The results indicate that corpus design and composition has a much greater influence on lexical variation than previously recognized, highlighting the need to evaluate internal representativeness in quantitative corpus-based research.